Research On Coded Modulation Technology Of Optical Fiber Communication System Based On Constellation Shaping

The rapid explosion of technologies and applications such as cloud computing,short video,autonomous driving and smart home has made daily life more convenient,but it has also led to a surge in panoramic traffic on the Internet.As one of the infrastructures providing underlying traffic carrying and technical support,the optical fiber communication system needs to develop towards higher rate and longer distance to cover the communication demands of users.However,since the transmit power cannot be too high due to the characteristics of optical fiber,thus the transmission distance without repeaters is limited.And if repeaters and photoelectric conversion are frequently utilized to extend reach,the transmission capacity will be reduced.Therefore,how to effectively improve the system capacity and transmission distance is the critical issue of optical fiber communication.For the above considerations,technologies such as constellation shaping and coded modulation are widely used,and become research hotspots in this field.On the theoretical basis of probabilistic shaping,this thesis studies the probabilistic shaping coded modulation scheme based on set partition,the probabilistic shaping coding modulation scheme based on concatenated trellis coded modulation and the probabilistic shaping decoding method based on neural network LLR(log likelihood ratio)calculation.The main research contents and innovations of this thesis are as follows:(1)In order to improve the transmission capacity and effective transmission distance of optical communication system,a probabilistic shaping coded modulation scheme based on set partition is proposed.In this scheme,signals in transmission constellation are allocated to different sets according to their energy,and the expected probability distributions are matched by adjusting the proportion of these sets to transmit.The shaping effect,bit error rate(BER)performance and compatibility with forward error-correcting coding are studied by simulation.The research results show that under the same shaping parameters,compared with the 16QAM probability shaping scheme using CCDM(Constant Composition Distribution Matching)distribution matcher,the proposed scheme achieves a signal-to-noise ratio gain of at least 2.5dB when it reaches the BER threshold of 10-3.(2)In order to improve the BER performance of the probability shaping scheme in trellis coded modulation,a cascaded trellis coded modulation probability shaping scheme based on subset re-mapping is proposed.The scheme uses CCDM to match the probability distribution,and further protects the bits both for subset selection and signal selection in the subset through the combination of concatenating RS codes with convolution codes and quadratic mapping processing of subsets.The antiinterference ability of the scheme is studied by simulation.The research results show that when the system bit error rate is 10-3,compared with the non-cascaded trellis coding modulation probability shaping scheme based on subset secondary mapping,there is more than 1dB improvement.The signal-to-noise ratio gain can effectively enhance the anti-interference ability of the system.(3)In order to reduce the decoding error of probabilistic shaping system,a decoding scheme based on neural network LLR calculation is proposed.In this scheme,a high-precision fitting of the LLR values required by the LLR BP(Belief Propagation)decoding algorithm is realized through neural network,and the ber performance is highly close to the LLR BP decoding algorithm based on the exact LLR value.The error performance and computational complexity of the scheme are simulated.The results show that the computational accuracy of the LLR algorithm based on neural network is more than 100 times higher than that of the LLR algorithm based on maximum logarithm approximation,and the SNR gain is obtained.